19th Asia-Pacific Vibration Conference (APVC2021)

247 / 2021-04-14 17:15:24

Howling abnormal sound diagnosis of aircraft based on spectrum visibility graph

Aircraft howling abnormal sound,Diagnosis,Visibility graph

Special Sessions > Applications of machine learning in vibration and noise problems

The howling abnormal sound of aircraft can affect the reliability of aircraft and the comfort of passengers. Identifying the fault types of the howling abnormal sound in time can provide an important reference for taking operation and making later maintenance. This paper presents a new method for this purpose. In this method, the frequency spectrum of the sound signal obtained by FFT is divided to the several sub-bands. The visibility graph is then constructed from the resulting sub-bands, where the considers the structural information of the spectrum. The KNN classifier is used to locate the aircraft howling abnormal sound, where the metric distance is calculated between graphs. The proposed method is investigated based on a real dataset collected by a certain type of aircraft, and experimental results demonstrate the priority and the great potential of the proposed method in real applications.